Research Of Foreign Substance Detection In Medicinal Liquid Based On Machine Vision

In recent years machine vision technology has developed very quickly and it's intelligence level has improved together with accuracy. Systems based on machine vision technology began to replace the human in many scenes of production and life. In some performance indicators, machine vision even performs better than the human visual perception. Liquid medical preparation is a widely used agents in clinical practice, and takes a important part in the pharmaceutical industry. Because of a variety of unavoidable reasons that lead foreign bodies introduced, which may deteriorate health, foreign body detection should be carried out during the production process to weed out substandard products, according to Chinese Pharmacopoeia. Currently artificial light inspection methods are widely used among domestic manufacturers, but the results are various over different people or different time, with low efficiency and high cost. So it is being replaced by automated machines light inspection. Liquid formulations detection technology based on machine vision could analysis products fast and accurate to ensure product quality, by analysis of liquid image sequence.This paper presents the definition of foreign body detection about Liquid formulations. It is thought that the importance and difficulty lies in the detection of small targets, if the quality of images are of high grade. This paper analyses current related systems and their feature. Then analyses the characteristics of foreign body in liquid pharmaceutical formulations and proposed a set of solutions. First, the design of the drive was seized liquid rotating machinery test platform. In a particular lighting system, the use of image acquisition equipment are foreign bodies moving image sequence. Study analyzes the foreign body in the bottle in the case of rotating the trajectory equation. Then the whole image needs to detect the liquid region to identify extraction. The picture in the background and analysis of the basic characteristics of the noise based on the proposed extraction and background noise suppression algorithms. Next image extracted area of foreign bodies may find several features of foreign body region, and with these features have proved a good cohesion and stability. Finally, the target feature vector, the target cluster, will belong to the same area of the image of foreign bodies of the many goals from the image sequence extracted by target motion characteristics, and ultimately determine whether the foreign body.Experimental platform through building, the use of 100 groups including foreign body image sequences to be tested. Experiments show that the proposed method in strong noise, can still reach 98% of the foreign matter detection rate, with good results.